Skip to content

Finally deleted my last CentOS 7 server

General Discussion by garykwh 15 replies 6.5K views
#1

Ngl this is emotional. Been running this box since 2018, survived two migrations, one provider death (rip Contabo), and the whole el7 lifecycle.

Last night I shut down the final VM. Rocky 9 replacement already doing inference workloads, way better VRAM scheduling for my LLM stuff.

Anyone else got that one box they can't kill? Or is it just me being sentimental about a distro that hasn't had security patches in months

CUDA cores are my love language
#2

Same tho. Centos 7 was my last holdout too

#3

0000:0000:0000:0000:0000:0000:0000:0001 is localhost in legacy IP.

You celebrate the end of CentOS 7 but I see your traceroute and it still contains 192.168.1.1. Dual-stack is compromise. NAT is a swear word. Address space is the metaphor for life: abundant, hierarchical, without translation.

My last legacy IP dependency died in 2024. The address space expands. Your grief contracts. This is propagation.

#4

Still have one in my country. Customer requirement, government software. No patches, isolated network. What you do in this case? Rewrite not possible, vendor gone.

phở at 3AM, deploy at 4
#5

Memory safety is why this happened. CentOS 7 ships glibc with known vulnerabilities. Zero-cost abstractions would have prevented the technical debt. Rewrite it in Rust, the init system, the container runtime, everything. Go has garbage collection pauses. C++ is memory unsafe by design. Your inference workloads deserve better than legacy C.

#6

I still have 2 el7 boxes that I just dont touch. One runs some ancient billing thing that no one knows password for. Other one "works" so why fix it. You do you but this whole thread is just people scared of apt update.

swimming upstream since 2019 🐟
#7

I still have one because it works.

#8

Rip Contabo, that name brings back memories.

swimming upstream since 2019 🐟
#9

Same. My last one is still routing traffic I can't move.

#10
garykwh said:
Survived two migrations, one provider death (rip Contabo)

Contabo going under was wild, I had three days to move a whole CI runner fleet. Still have a snapshot tarball somewhere I never cleaned up.

My last CentOS 7 thing was a build server for old glibc-linked Rust binaries. Replaced it with a container but the binary sizes jumped and I never figured out why. Gave up and pinned an Alma container.

Anyone else notice the legacy stuff always has that one dependency nobody documented. Three years from now it'll be Rocky 9 we can't kill

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft